Photovoltaic power station bird intrusion early warning method and system fused with multispectral detection

Abstract

The invention discloses a photovoltaic power station bird intrusion early warning method and system fused with multispectral detection, and relates to the technical field of intrusion early warning. The method comprises the following steps: carrying out detection coverage analysis on a target monitoring area, and carrying out multispectral detection network deployment according to an analysis result; bird key point fusion detection is carried out based on the multispectral data stream, and intrusion risk spatial features are output; according to the bird repelling device distribution and the invasion risk space position, bird repelling response delay is calculated and output; with the invasion risk space position as a starting point, bird invasion prediction is carried out, and bird repelling control parameters are output; matching real-time bird repelling early warning levels based on the bird repelling control parameters; and when bird invasion early warning is performed according to the real-time bird repelling early warning level, synchronously activating a bird repelling device to perform directional bird repelling by adopting the bird repelling control parameters. The technical problem that in the prior art, the accuracy and timeliness of bird intrusion detection of a photovoltaic power station are insufficient is solved, and the technical effect of improving the accuracy and timeliness of bird intrusion detection is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of intrusion warning, and in particular to a photovoltaic power station bird intrusion warning method and system integrating multi-spectral detection. Background Art

[0002] As the number and area of ​​photovoltaic power plants continue to grow, the challenges they face in their operations are becoming increasingly prominent. Bird intrusion has become a major challenge to the safe and stable operation, economic efficiency, and coordinated ecological and environmental development of photovoltaic power plants. Bird intrusion poses multiple risks to photovoltaic power plants. Physically, birds nesting and defecating on photovoltaic panels directly cover the panels with pollutants such as bird droppings. Bird droppings are highly corrosive, and long-term attachment can erode the protective coating on the panels, accelerating panel aging and reducing their photoelectric conversion efficiency, thereby impacting the power generation of the entire photovoltaic power plant. Furthermore, debris such as branches and weeds used by birds for nesting can be blown away in severe weather conditions such as strong winds, potentially colliding with photovoltaic panels or electrical equipment within the power plant, causing damage and other accidents, seriously threatening the safe operation of the photovoltaic power plant. Existing bird control technologies mostly rely on single-device monitoring, which suffers from incomplete monitoring coverage, slow response speed, and the inability to accurately locate bird activity. This results in missed and false alarms in early warning systems, and fails to meet the high-precision and timely requirements for bird intrusion warnings in photovoltaic power plants. Summary of the Invention

[0003] The present application provides a photovoltaic power station bird intrusion warning method and system integrating multi-spectral detection, which solves the technical problems of insufficient accuracy and timeliness of bird intrusion detection in photovoltaic power stations in the prior art.

[0004] In a first aspect of the present application, a method for early warning of bird intrusion in photovoltaic power plants integrating multispectral detection is provided, the method comprising: A detection coverage analysis is performed on the target monitoring area, and a multispectral detection network is deployed based on the analysis results, wherein the multispectral detection network includes a microwave radar array, a visible light camera array, and an infrared thermal imager array; bird key point fusion detection is performed based on the multispectral data stream collected by the multispectral detection network, and invasion risk spatial characteristics are output, wherein the invasion risk spatial characteristics include invasion risk spatial position, invasion risk speed, and invasion risk direction; based on the distribution of bird repellent devices and the invasion risk spatial position, a bird repellent response delay is calculated and output; starting from the invasion risk spatial position, bird invasion prediction is performed based on the invasion risk speed, invasion risk direction, and bird repellent response delay, and bird repellent control parameters are output; real-time bird repellent warning levels are matched based on the bird repellent control parameters; when a bird invasion warning is performed based on the real-time bird repellent warning level, the bird repellent control parameters are used to synchronously activate the bird repellent devices for directional bird repellent.

[0005] A second aspect of the present application provides a photovoltaic power station bird intrusion warning system integrating multi-spectral detection, the system comprising: A deployment module is used to perform detection coverage analysis on the target monitoring area and deploy a multispectral detection network based on the analysis results, wherein the multispectral detection network includes a microwave radar array, a visible light camera array and an infrared thermal imager array; a detection module is used to perform bird key point fusion detection based on the multispectral data stream collected by the multispectral detection network, and output invasion risk spatial characteristics, wherein the invasion risk spatial characteristics include invasion risk spatial position, invasion risk speed and invasion risk direction; a calculation module is used to calculate and output the bird repellent response delay based on the distribution of bird repellent devices and the invasion risk spatial position; a prediction module is used to predict bird invasion based on the invasion risk speed, invasion risk direction and bird repellent response delay with the invasion risk spatial position as the starting point, and output bird repellent control parameters; a matching module is used to match the real-time bird repellent warning level based on the bird repellent control parameters; a bird repellent module is used to use the bird repellent control parameters to synchronously activate the bird repellent device for directional bird repellent when issuing a bird invasion warning based on the real-time bird repellent warning level.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, a detection coverage analysis is performed on the target monitoring area. Based on the analysis results, a multispectral detection network is deployed. The multispectral detection network comprises a microwave radar array, a visible light camera array, and an infrared thermal imager array. Next, a fusion detection of bird key points is performed based on the multispectral data stream collected by the multispectral detection network. Intrusion risk spatial features are output, including intrusion risk spatial location, intrusion risk speed, and intrusion risk direction. Furthermore, the bird repellent response delay is calculated and output based on the distribution of bird repellent devices and the intrusion risk spatial location. Then, starting from the intrusion risk spatial location, bird intrusion prediction is performed based on the intrusion risk speed, intrusion risk direction, and bird repellent response delay, and bird repellent control parameters are output. Finally, the bird repellent control parameters are used to match the real-time bird repellent warning level. When a bird intrusion warning is issued based on the real-time bird repellent warning level, the bird repellent control parameters are used to synchronize and activate the bird repellent devices for targeted bird repelling. This method solves the technical problem of insufficient accuracy and timeliness in bird intrusion detection in photovoltaic power plants in the prior art, achieving the technical effect of improving the accuracy and timeliness of bird intrusion detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0008] Figure 1 A schematic flow chart of a method for early warning of bird intrusion in photovoltaic power stations using multi-spectral detection, provided in an embodiment of the present application; Figure 2 A schematic structural diagram of a photovoltaic power station bird intrusion warning system integrating multi-spectral detection provided in an embodiment of the present application.

[0009] Explanation of the accompanying drawings: deployment module 11, detection module 12, calculation module 13, prediction module 14, matching module 15, bird-repelling module 16. DETAILED DESCRIPTION

[0010] The present application solves the technical problem of insufficient accuracy and timeliness of bird intrusion detection in photovoltaic power stations in the prior art by providing a photovoltaic power station bird intrusion warning method and system integrating multi-spectral detection.

[0011] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0012] It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices.

[0013] Example 1, as Figure 1 As shown, the present application provides a photovoltaic power station bird intrusion warning method integrating multi-spectral detection, wherein the method includes: A detection coverage analysis is performed on the target monitoring area, and a multispectral detection network is deployed based on the analysis results, wherein the multispectral detection network includes a microwave radar array, a visible light camera array, and an infrared thermal imager array.

[0014] The system conducts a detection coverage analysis of the target monitoring area to determine the area's monitoring requirements, the target's dynamic behavior, and possible intrusion paths. By analyzing the target area's terrain, environment, and potential intrusion risks, the system assesses and determines the appropriate detection range and coverage requirements. Based on this analysis, the system deploys a multispectral detection network consisting of a microwave radar array, a visible light camera array, and an infrared thermal imager array. The microwave radar array is used for long-range detection in varying climatic conditions, capturing the target's location and trajectory; the visible light camera array provides high-resolution visual images to assist in identifying the target's specific morphology and behavioral characteristics; and the infrared thermal imager array captures the target's thermal radiation signature, particularly in low-light or nighttime conditions, to help identify potential intruders. Through the combined use of these three detection methods, the system achieves comprehensive coverage of the target monitoring area, acquiring timely, multi-dimensional detection data and providing accurate monitoring information for bird intrusion warnings.

[0015] Based on the multispectral data stream collected by the multispectral detection network, bird key point fusion detection is performed to output invasion risk spatial features, wherein the invasion risk spatial features include invasion risk spatial position, invasion risk speed and invasion risk direction.

[0016] The system uses a multispectral detection network (including microwave radar, visible light cameras, and infrared thermal imagers) to collect real-time multispectral data streams from the target area. By analyzing this data, the system can extract key bird information, such as flight path, location, and speed.

[0017] The system fuses data from a microwave radar array, a visible light camera array, and an infrared thermal imager array to identify bird flight paths and activity areas. This data fusion eliminates potential blind spots and errors associated with single detectors in different detection modes, providing more accurate bird location information. Based on this identified key point information, the system outputs intrusion risk spatial features, which include intrusion risk spatial position, intrusion risk speed, and intrusion risk direction. The intrusion risk spatial position represents the bird's position within the monitoring area, the intrusion risk speed represents the bird's flight speed, and the intrusion risk direction represents the bird's flight direction.

[0018] Furthermore, bird key point fusion detection is performed based on the multispectral data stream collected by the multispectral detection network, and invasion risk spatial features are output, including: The multispectral detection network collects microwave image sequence sets, visible light image sequence sets and infrared thermal imaging image sequence sets based on the microwave radar array, visible light camera array and infrared thermal imaging camera array, respectively, to form the multispectral data stream; performs continuous frame key point recognition on the microwave image sequence sets to obtain a flight continuous frame array corresponding to the microwave radar array; spatially fuses the flight continuous frame array to output a first-dimensional risk space feature; and similarly, outputs a second-dimensional risk space feature and a third-dimensional risk space feature by performing continuous frame key point recognition processing on the visible light image sequence set and the infrared thermal imaging image sequence set; performs bird key point fusion detection on the first-dimensional risk space feature, the second-dimensional risk space feature and the third-dimensional risk space feature to output the invasion risk space feature.

[0019] The multispectral detection network uses a microwave radar array, a visible light camera array, and an infrared thermal imager array to collect and generate microwave image sequences, visible light image sequences, and infrared thermal imager sequences, respectively, forming a multispectral data stream. Each image sequence represents a different type of detection information, covering a different spectral range, providing multi-dimensional data support for bird monitoring.

[0020] The system identifies key points in consecutive frames of the acquired microwave image sequence, identifies the bird's flight path, and generates a flight continuous frame array corresponding to the microwave radar array. Each frame in the flight continuous frame array represents the bird's position and state at a specific moment. The system fuses the key point information from multiple frames through temporal and spatial alignment, eliminating temporal and spatial errors to obtain the bird's trajectory. This fused data generates a first-dimensional risk space feature, which describes the bird's spatial position changes, flight trajectory, and corresponding velocity information within the monitoring area.

[0021] The system similarly identifies keypoints in consecutive frames for the collected visible light and infrared thermal image sequences. For each image sequence, the system converts the identified bird keypoint information into second- and third-dimensional risk space features to further characterize the bird's spatial behavior. The system then fuses these first-, second-, and third-dimensional risk space features, combining microwave radar, visible light, and infrared thermal imaging data to comprehensively analyze the bird's movement trajectories and behavioral patterns, ultimately outputting a final invasion risk space signature.

[0022] Furthermore, performing bird key point fusion detection on the first-dimensional risk space feature, the second-dimensional risk space feature, and the third-dimensional risk space feature to output the invasion risk space feature includes: The target monitoring area and the acquisition window of the multispectral data stream are used as dual limiting conditions to retrieve real-time regional environmental characteristics online; modal weight matching is performed on the real-time regional environmental characteristics to obtain multispectral trust level weights; after confidence verification of the first-dimensional risk space characteristics, the second-dimensional risk space characteristics and the third-dimensional risk space characteristics according to the multispectral trust level weights, bird key point fusion detection is performed to output the invasion risk space characteristics.

[0023] Specifically, the target monitoring area and the acquisition window of the multispectral data stream are used as dual constraints. These constraints are used to retrieve real-time regional environmental characteristics online, acquiring environmental data for the area, including factors such as weather, light, and temperature. Next, the acquired real-time regional environmental characteristics are subjected to modal weight matching to obtain multispectral trust level weights. Specifically, the system performs modal weight matching on the real-time regional environmental characteristics collected. This means that each environmental data point is assigned a weight value based on factors such as the data source, accuracy, and environmental conditions. These weights reflect the contribution of each data point to overall environmental monitoring. Through modal weight matching, the system combines the weights of different data sources to generate multispectral trust level weights. These multispectral trust level weights are used to measure the credibility of multispectral detection data and provide a basis for the subsequent calculation and fusion of invasion risk spatial features. Data with higher trust level weights have a greater impact on the final judgment, thereby improving the system's accuracy in bird invasion monitoring and prediction.

[0024] Based on the multispectral trust level weights, the system performs confidence checks on the first-, second-, and third-dimensional risk space features. Specifically, the system verifies the risk space features of each dimension using the obtained multispectral trust level weights to confirm their credibility and validity. This weighted verification ensures that low-quality data will not affect the final results.

[0025] After completing the confidence verification, the system performs bird key point fusion detection, fuses the first-dimensional risk space features, the second-dimensional risk space features, and the third-dimensional risk space features to obtain the invasion risk space features.

[0026] Furthermore, performing continuous frame key point recognition on the microwave image sequence set to obtain a flight continuous frame array corresponding to the microwave radar array includes: Point cloud density aggregation of bird key points is performed based on historical microwave image data, and a bird head density threshold and a bird wing flapping density threshold are output; the bird head density threshold is used to traverse the first microwave image frame of the first microwave radar node, and the first head area is screened and identified; with the first head area as the starting point, the bird wing flapping density threshold is used to identify symmetrically distributed point cloud areas, and the first wing flapping symmetric area is screened and identified; based on the characteristics of bird flight behavior, the first head area and the first wing flapping symmetric area are connected, and a first frame of bird recognition result is output; and similarly, continuous frame key point recognition is performed on the first microwave image sequence of the first microwave radar node, and a first flight continuous frame including multiple frames of bird recognition results is output; and similarly, continuous frame key point recognition is performed on the microwave image sequence set to obtain the flight continuous frame array.

[0027] The system analyzes the key point data of birds in historical microwave images and aggregates the point cloud density of each key point. It also counts the distribution of key points of birds during flight (such as the bird's head and wing flapping parts) in microwave images to identify the density of these features in microwave images. The mode of these densities is then selected to output the bird's head density threshold and bird's wing flapping density threshold. The bird's head density threshold is used to identify the bird's head features, and the bird's wing flapping density threshold is used to determine the motion characteristics of the bird's wing flapping.

[0028] The system uses the bird head density threshold to traverse the first microwave image frame of the first microwave radar node and screen and identify the first head region. Specifically, the system analyzes each frame pixel by pixel and compares it with the bird head density threshold to identify areas that meet head characteristics and mark them as the first head region. Starting from the identified first head region, the system further uses the bird wingbeat density threshold to identify symmetrically distributed point cloud regions, screening and identifying the first wingbeat symmetrical region.

[0029] Based on the characteristics of bird flight behavior, the system connects the first head region and the first symmetrical wingbeat region, outputting the first frame of bird recognition results to form complete flight trajectory data. Similarly, the system performs continuous frame keypoint recognition on the first microwave image sequence of the first microwave radar node, outputting the first continuous flight frame, which includes multiple frames of bird recognition results. Finally, the system uses the same method to perform continuous frame keypoint recognition on the microwave image sequence set, ultimately generating an array of continuous flight frames.

[0030] Furthermore, the flight continuous frame array is spatially fused to output a first-dimensional risk space feature, including: After time-series alignment of the flight continuous frame array, the flight continuous frame array is spatially aligned and fused according to the spatial layout characteristics of the microwave radar array, and a spatial bird frame coordinate sequence is output; the tail frame coordinates are extracted from the spatial bird frame coordinate sequence to obtain the first-dimensional risk space position; the adjacent frame displacement of the spatial bird frame coordinate sequence is calculated to output the first-dimensional risk speed; the displacement vector of the spatial bird frame coordinate sequence is calculated to output the first-dimensional risk direction, wherein the first-dimensional risk space position, the first-dimensional risk speed and the first-dimensional risk direction constitute the first-dimensional risk space characteristics.

[0031] The system performs temporal alignment on the continuous flight frame array, ensuring that the frame data in the image sequence is arranged in chronological order. Then, based on the spatial characteristics of the microwave radar array layout, the system performs spatial alignment and fusion on the continuous flight frame array. By uniformly transforming the spatial coordinates of different frames, the system ensures that the bird position data in each frame of the image accurately corresponds to the same spatial coordinate system, thereby generating a precise spatial bird frame coordinate sequence.

[0032] After completing temporal and spatial alignment, the system extracts the tail frame coordinates from the spatial bird frame coordinate sequence to obtain the first-dimensional risk space position, which reflects the bird's specific location within the monitoring area. The system calculates the displacement of adjacent frames in the spatial bird frame coordinate sequence and outputs the first-dimensional risk velocity. By calculating the displacement of the bird between adjacent frames, the system can assess the bird's movement speed, providing key data for subsequent invasion prediction and prevention. Furthermore, the system calculates the displacement vector of the spatial bird frame coordinate sequence and outputs the first-dimensional risk direction. By analyzing the bird's movement trajectory, the system determines the bird's flight direction, providing directional information for the early warning system. Ultimately, the resulting first-dimensional risk space position, first-dimensional risk velocity, and first-dimensional risk direction constitute the first-dimensional risk space characteristics.

[0033] Furthermore, after performing confidence verification on the first-dimensional risk space feature, the second-dimensional risk space feature, and the third-dimensional risk space feature according to the multispectral trust level weight, bird key point fusion detection is performed to output the invasion risk space feature, including: Extracting a trimodal risk space position from the first-dimensional risk space features, the second-dimensional risk space features, and the third-dimensional risk space features, wherein the trimodal risk space position includes the first-dimensional risk space position, the second-dimensional risk space position, and the third-dimensional risk space position; fusing the trimodal risk space positions based on the multispectral trust level weights, and outputting the weighted centroid as the intrusion risk space position; extracting trimodal risk speed and trimodal risk direction from the first-dimensional risk space features, the second-dimensional risk space features, and the third-dimensional risk space features; performing dynamic error balance on the trimodal risk speed based on the multispectral trust level weights, and outputting the intrusion risk speed; performing directional accuracy optimization on the trimodal risk speed based on the multispectral trust level weights, and outputting the intrusion risk direction.

[0034] The system extracts trimodal risk space locations from the first, second, and third-dimensional risk space features. Each risk space location represents the location of a bird within the monitoring area under different spectra. The system uses this location data to accurately locate bird intrusion risk areas.

[0035] The system performs weighted fusion on the three-modal risk spatial positions according to the multispectral trust level weights to obtain a weighted centroid, which represents the core position of the bird invasion risk, that is, the invasion risk spatial position.

[0036] After completing spatial position fusion, the system extracts trimodal risk speed and trimodal risk direction from the first-dimensional risk space characteristics, the second-dimensional risk space characteristics and the third-dimensional risk space characteristics; based on the multispectral trust level weight, the trimodal risk speed is dynamically error balanced and the intrusion risk speed is output; based on the multispectral trust level weight, the direction accuracy of the trimodal risk speed is optimized and the intrusion risk direction is output.

[0037] According to the distribution of the bird repellent devices and the spatial location of the invasion risk, a bird repellent response delay is calculated and output.

[0038] Based on known bird repellent device distribution information, the system determines the distance between each device and the intrusion risk location. By obtaining the device's geographic location and the intrusion risk location, the system calculates the effective distance from each device to the risk location. Based on this, the system calculates the speed of sound waves in air and the distance between the device and the intrusion risk location to determine the response delay of each device. The response delay is the time it takes for the device to detect a bird target at the intrusion risk location and then initiate a response signal.

[0039] Furthermore, according to the distribution of the bird repellent devices and the spatial location of the intrusion risk, the bird repellent response delay is calculated and outputted, including: The device distance distribution is calculated based on the bird-repellent device distribution and the intrusion risk spatial position; P effective devices are screened out from the device distance distribution based on the effective intrusion protection distance; sound wave propagation delays are calculated based on P effective distances between the P effective devices and the intrusion risk spatial position, and P parameter adjustment response delays are output; after adding P device startup delays and P parameter adjustment response delays of the P effective devices, the maximum value of the sum is extracted as the bird-repellent response delay.

[0040] By calculating the distance between each bird-repellent device and the invasion risk spatial location, a device distance distribution map is obtained, which shows the relative position relationship of each bird-repellent device relative to the invasion risk spatial location.

[0041] Based on the effective intrusion protection distance, the system selects P effective devices from the device distance distribution. The effective intrusion protection distance refers to the area effectively covered by each bird repellent device. Based on the bird's flight path and the predetermined protection requirements, P effective bird repellent devices are selected within this range. The system calculates the sound wave propagation delay based on the P effective distances between the P effective devices and the intrusion risk spatial location. The system calculates the time required for the sound wave to propagate from each effective bird repellent device to the intrusion risk spatial location and outputs P parameter response delays. Each parameter response delay reflects the time from the activation of the bird repellent device to the sound wave signal reaching the intrusion target. The system sums the P device startup delays and the P parameter response delays of the P effective devices and extracts the maximum value of the sum as the bird repellent response delay. The device startup delay refers to the time from the receipt of the start signal to the actual activation of the bird repellent device. By summing these two delay values, the system ultimately outputs a combined bird repellent response delay, which is the maximum response delay among all effective devices.

[0042] Taking the invasion risk spatial position as a starting point, bird invasion prediction is performed according to the invasion risk speed, invasion risk direction and bird repelling response delay, and bird repelling control parameters are output.

[0043] The system uses the spatial location of the intrusion risk as the starting point for bird intrusion prediction. This location represents the bird's current position within the monitored area. Based on this starting point, the system calculates the bird's future trajectory, combining the intrusion risk speed and intrusion risk direction, to determine the bird's intended arrival location. Based on the intended location and the bird repellent response delay, the system calculates bird repellent control parameters to provide decision support for the activation of the bird repellent device. These control parameters include key operating parameters such as the device's activation time, direction, and intensity. This ensures that the device can respond promptly and accurately to the path and location of intruding birds, achieving effective bird protection.

[0044] Furthermore, taking the invasion risk spatial position as the starting point, bird invasion prediction is performed according to the invasion risk speed, invasion risk direction and bird repelling response delay, and bird repelling control parameters are output, including: The flight inertia is predicted based on the intrusion risk speed, the invasion risk direction and the bird-repelling response delay, and a flight inertia virtual vector is output; the flight inertia virtual vector is fitted to the intrusion risk spatial position to locate the invasion prediction spatial position; based on the intrusion prediction spatial position and P bird-repelling distances of P effective devices, a target device corresponding to the minimum positioning distance is screened from the P effective devices; and the bird-repelling control parameters are reversely matched based on the device position characteristics of the target device and the intrusion prediction spatial position.

[0045] The system performs flight inertia prediction based on the intrusion risk speed and direction. This prediction analyzes the bird's speed and direction, calculates its inertial motion characteristics, and outputs a virtual flight inertia vector. This virtual flight inertia vector represents the bird's dynamic characteristics and helps the system predict the bird's future movement trends. The system then fits the virtual flight inertia vector to the intrusion risk spatial location to determine the predicted intrusion spatial location. This fitting process enables the system to accurately predict the bird's future location based on its inertial trajectory and current location. The system then selects the bird repellent device closest to the predicted intrusion spatial location based on the predicted intrusion spatial location and the P distances between P effective bird repellent devices. By calculating the distance between each effective device and the predicted intrusion spatial location, the system selects the most appropriate target device that will most effectively respond to bird intrusions. Finally, the system performs reverse matching based on the device location characteristics of the selected target device and the predicted intrusion spatial location to determine the bird repellent control parameters, including the activation time, direction, and intensity of the device. This ensures that the device activates promptly and accurately to respond to bird intrusions, maximizing prevention and control effectiveness.

[0046] Furthermore, the bird-repelling control parameters are matched inversely according to the device location characteristics of the target device and the intrusion prediction spatial location, including: Interactively obtain multiple sample bird repellent control information, wherein the sample bird repellent control information includes multiple sample device locations, multiple sample intrusion locations and multiple sample control parameters; store the multiple sample bird repellent control information based on the knowledge graph association to obtain a bird repellent control matching library; input the device location features and the intrusion prediction spatial location into the bird repellent control matching library, and reversely match the bird repellent control parameters.

[0047] The system interactively obtains multiple sample bird repellent control information, including multiple sample device locations, multiple sample intrusion locations, and multiple sample control parameters. This sample data can be obtained from historical monitoring data or field test data, covering different bird repellent device settings, intrusion event locations, and corresponding control parameters.

[0048] Based on a knowledge graph, the system associates and stores multiple bird repellent control information samples to construct a bird repellent control matching library. A knowledge graph is a structured data storage method that associates different bird repellent devices, intrusion locations, and control parameters, establishing a complex network of relationships. The system inputs device location characteristics and predicted intrusion spatial locations into the bird repellent control matching library. Using an inverse algorithm, the system performs matching and determines the most appropriate bird repellent control parameters.

[0049] A real-time bird-repelling warning level is matched based on the bird-repelling control parameters.

[0050] The system assesses the current invasion risk level based on the matched bird repellent control parameters (including the location, activation time, intensity and direction of the bird repellent device); using the bird repellent control parameters, it assesses the current bird invasion risk through a weighted algorithm and calculates the real-time bird repellent warning level.

[0051] For example, if the flock of birds is moving slowly, the estimated entry time is long, and the bird repellent device is close, the system may assess it as a low-risk warning level, which means that the activation response of the bird repellent device can be appropriately delayed and the intensity can be set to medium. If the flock of birds is moving quickly and is expected to enter the power plant's monitoring area in a short time, and the distance between the bird repellent device and the flock is far, the system will assess it as a medium-risk warning level, and it will be necessary to appropriately increase the intensity of the bird repellent device and the lead time for activation. If the flock of birds approaches quickly and is expected to enter the power plant in a very short time, the system will assess it as a high-risk warning level.

[0052] When a bird invasion warning is performed according to the real-time bird-repelling warning level, the bird-repelling control parameters are used to synchronously activate the bird-repelling device to perform directional bird-repelling.

[0053] The system assesses the urgency of bird intrusion based on the real-time bird repellent warning level. Based on different bird repellent warning levels, the system uses different bird repellent control parameters and synchronously activates bird repellent for targeted bird repellent.

[0054] In summary, the embodiments of the present application have at least the following technical effects: First, a detection coverage analysis is performed on the target monitoring area. Based on the analysis results, a multispectral detection network is deployed. The multispectral detection network comprises a microwave radar array, a visible light camera array, and an infrared thermal imager array. Next, a fusion detection of bird key points is performed based on the multispectral data stream collected by the multispectral detection network. Intrusion risk spatial features are output, including intrusion risk spatial location, intrusion risk speed, and intrusion risk direction. Furthermore, the bird repellent response delay is calculated and output based on the distribution of bird repellent devices and the intrusion risk spatial location. Then, starting from the intrusion risk spatial location, bird intrusion prediction is performed based on the intrusion risk speed, intrusion risk direction, and bird repellent response delay, and bird repellent control parameters are output. Finally, the bird repellent control parameters are used to match the real-time bird repellent warning level. When a bird intrusion warning is issued based on the real-time bird repellent warning level, the bird repellent control parameters are used to synchronize and activate the bird repellent devices for targeted bird repelling. This method solves the technical problem of insufficient accuracy and timeliness in bird intrusion detection in photovoltaic power plants in the prior art, achieving the technical effect of improving the accuracy and timeliness of bird intrusion detection.

[0055] Example 2 is based on the same inventive concept as the above-mentioned embodiment of a photovoltaic power station bird invasion early warning method integrating multi-spectral detection, such as Figure 2 As shown, the present application provides a photovoltaic power station bird intrusion warning system integrating multi-spectral detection, wherein the system includes: A deployment module 11 is configured to perform detection coverage analysis on the target monitoring area and deploy a multispectral detection network based on the analysis results, wherein the multispectral detection network includes a microwave radar array, a visible light camera array, and an infrared thermal imager array; a detection module 12 is configured to perform bird key point fusion detection based on the multispectral data stream collected by the multispectral detection network, and output invasion risk spatial features, wherein the invasion risk spatial features include invasion risk spatial position, invasion risk speed, and invasion risk direction; a calculation module 13 is configured to calculate and output a bird repellent response delay based on the distribution of bird repellent devices and the invasion risk spatial position; a prediction module 14 is configured to perform bird invasion prediction based on the invasion risk spatial position as the starting point, the invasion risk speed, the invasion risk direction, and the bird repellent response delay, and output bird repellent control parameters; a matching module 15 is configured to match a real-time bird repellent warning level based on the bird repellent control parameters; and a bird repellent module 16 is configured to, when issuing a bird invasion warning based on the real-time bird repellent warning level, synchronously activate the bird repellent device using the bird repellent control parameters for directional bird repellent.

[0056] Furthermore, the detection module 12 is configured to perform the following method: The multispectral detection network collects microwave image sequence sets, visible light image sequence sets and infrared thermal imaging image sequence sets based on the microwave radar array, visible light camera array and infrared thermal imaging camera array, respectively, to form the multispectral data stream; performs continuous frame key point recognition on the microwave image sequence sets to obtain a flight continuous frame array corresponding to the microwave radar array; spatially fuses the flight continuous frame array to output a first-dimensional risk space feature; and similarly, outputs a second-dimensional risk space feature and a third-dimensional risk space feature by performing continuous frame key point recognition processing on the visible light image sequence set and the infrared thermal imaging image sequence set; performs bird key point fusion detection on the first-dimensional risk space feature, the second-dimensional risk space feature and the third-dimensional risk space feature to output the invasion risk space feature.

[0057] Furthermore, the detection module 12 is configured to perform the following method: The target monitoring area and the acquisition window of the multispectral data stream are used as dual limiting conditions to retrieve real-time regional environmental characteristics online; modal weight matching is performed on the real-time regional environmental characteristics to obtain multispectral trust level weights; after confidence verification of the first-dimensional risk space characteristics, the second-dimensional risk space characteristics and the third-dimensional risk space characteristics according to the multispectral trust level weights, bird key point fusion detection is performed to output the invasion risk space characteristics.

[0058] Furthermore, the detection module 12 is configured to perform the following method: Point cloud density aggregation of bird key points is performed based on historical microwave image data, and a bird head density threshold and a bird wing flapping density threshold are output; the bird head density threshold is used to traverse the first microwave image frame of the first microwave radar node, and the first head area is screened and identified; with the first head area as the starting point, the bird wing flapping density threshold is used to identify symmetrically distributed point cloud areas, and the first wing flapping symmetric area is screened and identified; based on the characteristics of bird flight behavior, the first head area and the first wing flapping symmetric area are connected, and a first frame of bird recognition result is output; and similarly, continuous frame key point recognition is performed on the first microwave image sequence of the first microwave radar node, and a first flight continuous frame including multiple frames of bird recognition results is output; and similarly, continuous frame key point recognition is performed on the microwave image sequence set to obtain the flight continuous frame array.

[0059] Furthermore, the detection module 12 is configured to perform the following method: After time-series alignment of the flight continuous frame array, the flight continuous frame array is spatially aligned and fused according to the spatial layout characteristics of the microwave radar array, and a spatial bird frame coordinate sequence is output; the tail frame coordinates are extracted from the spatial bird frame coordinate sequence to obtain the first-dimensional risk space position; the adjacent frame displacement of the spatial bird frame coordinate sequence is calculated to output the first-dimensional risk speed; the displacement vector of the spatial bird frame coordinate sequence is calculated to output the first-dimensional risk direction, wherein the first-dimensional risk space position, the first-dimensional risk speed and the first-dimensional risk direction constitute the first-dimensional risk space characteristics.

[0060] Furthermore, the detection module 12 is configured to perform the following method: Extracting a trimodal risk space position from the first-dimensional risk space features, the second-dimensional risk space features, and the third-dimensional risk space features, wherein the trimodal risk space position includes the first-dimensional risk space position, the second-dimensional risk space position, and the third-dimensional risk space position; fusing the trimodal risk space positions based on the multispectral trust level weights, and outputting the weighted centroid as the intrusion risk space position; extracting trimodal risk speed and trimodal risk direction from the first-dimensional risk space features, the second-dimensional risk space features, and the third-dimensional risk space features; performing dynamic error balance on the trimodal risk speed based on the multispectral trust level weights, and outputting the intrusion risk speed; performing directional accuracy optimization on the trimodal risk speed based on the multispectral trust level weights, and outputting the intrusion risk direction.

[0061] Furthermore, the calculation module 13 is used to perform the following method: The device distance distribution is calculated based on the bird-repellent device distribution and the intrusion risk spatial position; P effective devices are screened out from the device distance distribution based on the effective intrusion protection distance; sound wave propagation delays are calculated based on P effective distances between the P effective devices and the intrusion risk spatial position, and P parameter adjustment response delays are output; after adding P device startup delays and P parameter adjustment response delays of the P effective devices, the maximum value of the sum is extracted as the bird-repellent response delay.

[0062] Furthermore, the prediction module 14 is configured to perform the following method: The flight inertia is predicted based on the intrusion risk speed, the invasion risk direction and the bird-repelling response delay, and a flight inertia virtual vector is output; the flight inertia virtual vector is fitted to the intrusion risk spatial position to locate the invasion prediction spatial position; based on the intrusion prediction spatial position and P bird-repelling distances of P effective devices, a target device corresponding to the minimum positioning distance is screened from the P effective devices; and the bird-repelling control parameters are reversely matched based on the device position characteristics of the target device and the intrusion prediction spatial position.

[0063] Furthermore, the prediction module 14 is configured to perform the following method: Interactively obtain multiple sample bird repellent control information, wherein the sample bird repellent control information includes multiple sample device locations, multiple sample intrusion locations and multiple sample control parameters; store the multiple sample bird repellent control information based on the knowledge graph association to obtain a bird repellent control matching library; input the device location features and the intrusion prediction spatial location into the bird repellent control matching library, and reversely match the bird repellent control parameters.

[0064] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0065] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

[0066] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. A photovoltaic power station bird invasion early warning method integrating multi-spectral detection, characterized in that: The method comprises: Conducting a detection coverage analysis of the target monitoring area and deploying a multispectral detection network based on the analysis results, wherein the multispectral detection network includes a microwave radar array, a visible light camera array, and an infrared thermal imager array; Performing bird key point fusion detection based on the multispectral data stream collected by the multispectral detection network, and outputting invasion risk spatial features, wherein the invasion risk spatial features include invasion risk spatial position, invasion risk speed, and invasion risk direction; Calculating and outputting a bird repellent response delay based on the distribution of the bird repellent device and the spatial location of the invasion risk; Taking the invasion risk spatial position as a starting point, predicting bird invasion according to the invasion risk speed, invasion risk direction and bird repelling response delay, and outputting bird repelling control parameters; Matching a real-time bird-repelling warning level based on the bird-repelling control parameters; When a bird invasion warning is performed according to the real-time bird-repelling warning level, the bird-repelling control parameters are used to synchronously activate the bird-repelling device to perform directional bird-repelling.

2. The photovoltaic power station bird invasion early warning method integrating multi-spectral detection as claimed in claim 1, characterized in that: Based on the multispectral data stream collected by the multispectral detection network, bird key point fusion detection is performed to output invasion risk spatial features, and the method includes: The multispectral detection network collects microwave image sequence sets, visible light image sequence sets and infrared thermal imaging image sequence sets based on the microwave radar array, visible light camera array and infrared thermal imager array, respectively, to form the multispectral data stream; Performing continuous frame key point recognition on the microwave image sequence set to obtain a flight continuous frame array corresponding to the microwave radar array; spatially fusing the flight continuous frame array to output a first-dimensional risk space feature; Similarly, by performing continuous frame key point recognition processing on the visible light image sequence set and the infrared thermal imaging image sequence set, a second-dimensional risk space feature and a third-dimensional risk space feature are output; Perform bird key point fusion detection on the first-dimensional risk space feature, the second-dimensional risk space feature, and the third-dimensional risk space feature, and output the invasion risk space feature.

3. The photovoltaic power station bird invasion early warning method integrating multi-spectral detection as claimed in claim 2, characterized in that: Performing bird key point fusion detection on the first-dimensional risk space feature, the second-dimensional risk space feature, and the third-dimensional risk space feature to output the invasion risk space feature, the method comprising: Using the target monitoring area and the acquisition window of the multispectral data stream as dual limiting conditions, online retrieval of real-time regional environmental characteristics; Performing modal weight matching on the real-time regional environmental features to obtain a multispectral trust level weight; After confidence verification is performed on the first-dimensional risk space feature, the second-dimensional risk space feature, and the third-dimensional risk space feature according to the multispectral trust level weight, bird key point fusion detection is performed to output the invasion risk space feature.

4. The photovoltaic power station bird invasion early warning method integrating multi-spectral detection as claimed in claim 2, characterized in that: Performing continuous frame key point recognition on the microwave image sequence set to obtain a flight continuous frame array corresponding to the microwave radar array, the method comprising: Based on historical microwave image data, the point cloud density of bird key points is aggregated to output the bird head density threshold and the bird wing flapping density threshold; Using the bird head density threshold to traverse the first microwave image frame of the first microwave radar node, and screening and identifying the first head area; Taking the first head region as a starting point, using the bird wingbeat density threshold to perform symmetrically distributed point cloud region recognition, and screening and identifying the first wingbeat symmetrical region; Based on the bird's flight behavior characteristics, connecting the first head region and the first wing-flapping symmetrical region, and outputting a first frame of bird recognition results; Similarly, performing continuous frame key point recognition on the first microwave image sequence of the first microwave radar node, and outputting a first flight continuous frame including multiple frames of bird recognition results; Similarly, continuous frame key point recognition is performed on the microwave image sequence set to obtain the flight continuous frame array.

5. The photovoltaic power station bird invasion early warning method integrating multi-spectral detection as claimed in claim 4, characterized in that: Spatially fusing the flight continuous frame array to output a first-dimensional risk space feature, the method comprising: After the time sequence alignment of the flight continuous frame array, the flight continuous frame array is spatially aligned and fused according to the spatial characteristics of the microwave radar array, and a spatial bird frame coordinate sequence is output; Extracting tail frame coordinates from the spatial bird frame coordinate sequence to obtain a first-dimensional risk space position; performing adjacent frame displacement calculation on the spatial bird frame coordinate sequence and outputting a first-dimensional risk velocity; A displacement vector is calculated for the spatial bird frame coordinate sequence to output a first-dimensional risk direction, wherein the first-dimensional risk space position, the first-dimensional risk speed, and the first-dimensional risk direction constitute the first-dimensional risk space feature.

6. The photovoltaic power station bird invasion early warning method integrating multi-spectral detection as claimed in claim 3, characterized in that: After performing confidence verification on the first-dimensional risk space feature, the second-dimensional risk space feature, and the third-dimensional risk space feature according to the multispectral trust level weight, performing bird key point fusion detection, and outputting the invasion risk space feature, the method includes: Extracting a trimodal risk space position from the first-dimensional risk space feature, the second-dimensional risk space feature, and the third-dimensional risk space feature, wherein the trimodal risk space position includes the first-dimensional risk space position, the second-dimensional risk space position, and the third-dimensional risk space position; According to the multispectral trust level weight, the three-modal risk space positions are integrated, and a weighted centroid is output as the intrusion risk space position; Extracting trimodal risk speed and trimodal risk direction from the first-dimensional risk space characteristics, the second-dimensional risk space characteristics, and the third-dimensional risk space characteristics; Performing dynamic error balancing on the three-modal risk speeds according to the multispectral trust level weights, and outputting the intrusion risk speed; The direction accuracy of the three-modal risk speed is optimized according to the multispectral trust level weight, and the intrusion risk direction is output.

7. The photovoltaic power station bird invasion early warning method integrating multi-spectral detection as claimed in claim 1, characterized in that: Calculating and outputting a bird-repelling response delay based on the distribution of the bird-repelling devices and the intrusion risk spatial location, the method comprising: Calculating the distance distribution of the devices based on the distribution of the bird repellent devices and the spatial location of the intrusion risk; Filtering P effective devices from the device distance distribution according to the effective intrusion protection distance; Calculate the acoustic wave propagation delay based on the P effective distances between the P effective devices and the intrusion risk space location, and output P parameter adjustment response delays; After adding up the P device startup delays and the P parameter adjustment response delays of the P valid devices, the maximum value of the sum is extracted as the bird-repelling response delay.

8. The photovoltaic power station bird invasion early warning method integrating multi-spectral detection as claimed in claim 7, characterized in that: Taking the invasion risk spatial position as a starting point, predicting bird invasion based on the invasion risk speed, invasion risk direction, and bird repelling response delay, and outputting bird repelling control parameters, the method includes: Perform flight inertia prediction based on the invasion risk speed, invasion risk direction, and bird repelling response delay, and output a flight inertia virtual vector; Fitting the flight inertia virtual vector to the invasion risk spatial position to locate the invasion prediction spatial position; According to the predicted intrusion spatial position and the P bird-repelling distances of the P effective devices, a target device corresponding to a minimum positioning distance value is selected from the P effective devices; The bird-repelling control parameters are reversely matched according to the device location characteristics of the target device and the intrusion prediction spatial location.

9. The photovoltaic power station bird invasion early warning method integrating multi-spectral detection as claimed in claim 8, characterized in that: Inversely matching the bird-repelling control parameters according to the device location characteristics of the target device and the intrusion prediction spatial location, the method includes: interactively obtaining a plurality of sample bird-repelling control information, wherein the sample bird-repelling control information includes a plurality of sample device positions, a plurality of sample intrusion positions, and a plurality of sample control parameters; The plurality of sample bird repellent control information is stored in association with each other based on the knowledge graph to obtain a bird repellent control matching library; The device location characteristics and the intrusion prediction spatial location are input into the bird-repelling control matching library, and the bird-repelling control parameters are matched in reverse order.

10. A photovoltaic power station bird invasion warning system integrating multi-spectral detection, characterized in that: A method for early warning of bird intrusion in photovoltaic power stations by integrating multi-spectral detection according to any one of claims 1 to 9, the system comprising: A deployment module is used to perform detection coverage analysis on the target monitoring area and deploy a multispectral detection network based on the analysis results, wherein the multispectral detection network includes a microwave radar array, a visible light camera array, and an infrared thermal imager array; a detection module, configured to perform bird key point fusion detection based on the multispectral data stream collected by the multispectral detection network, and output invasion risk spatial features, wherein the invasion risk spatial features include invasion risk spatial position, invasion risk speed, and invasion risk direction; a calculation module, configured to calculate and output a bird-repelling response delay based on the distribution of the bird-repelling devices and the spatial location of the invasion risk; A prediction module is configured to use the invasion risk spatial position as a starting point, perform bird invasion prediction based on the invasion risk speed, invasion risk direction, and bird repelling response delay, and output bird repelling control parameters; A matching module, configured to match a real-time bird-repelling warning level based on the bird-repelling control parameters; The bird-repelling module is used to synchronously activate the bird-repelling device to carry out directional bird-repelling by using the bird-repelling control parameters when a bird invasion warning is issued according to the real-time bird-repelling warning level.

Images